Using Campaign GPS Data to Model Slip Rates on the Alpine Fault

New Zealand Journal of Geology and Geophysics

ISSN: 0028-8306 (Print) 1175-8791 (Online) Journal homepage: http://www.tandfonline.com/loi/tnzg20

A geodetic study of the Alpine Fault through South Westland: using campaign GPS data to model slip rates on the Alpine Fault

Chris J. Page, Paul H. Denys & Chris F. Pearson

To cite this article: Chris J. Page, Paul H. Denys & Chris F. Pearson (2018): A geodetic study of the Alpine Fault through South Westland: using campaign GPS data to model slip rates on the Alpine Fault, New Zealand Journal of Geology and Geophysics, DOI: 10.1080/00288306.2018.1494006 To link to this article: https://doi.org/10.1080/00288306.2018.1494006

View supplementary material

Published online: 09 Aug 2018.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=tnzg20 NEW ZEALAND JOURNAL OF GEOLOGY AND GEOPHYSICS https://doi.org/10.1080/00288306.2018.1494006

RESEARCH ARTICLE A geodetic study of the Alpine Fault through South Westland: using campaign GPS data to model slip rates on the Alpine Fault Chris J. Page, Paul H. Denys and Chris F. Pearson School of Surveying, University of Otago, Dunedin, New Zealand

ABSTRACT ARTICLE HISTORY Although the Alpine Fault has been studied extensively, there have been few geodetic studies Received 11 December 2017 in South Westland. We include a series of new geodetic measurements from sites across the Accepted 25 June 2018 Haast Pass and preliminary results from a recently established network, the Cascade array KEYWORDS that extends from the Arawhata River to Lake McKerrow, a region that previously had few Alpine Fault; slip rates; South geodetic measurements. We compare the slip rates based on models that include both Westland; deformation; GPS single and double faults, and consider Alpine Fault dips of 55° and near vertical. Our preferred solution models the Alpine Fault as an inﬁnitely long fault, dipping at 55° with a second (proxy) fault to account for (inboard) distributed deformation. This gives results that are consistent with the Alpine Fault being a predominantly strike-slip fault with a slip rate of 30 ± 2 mm/yr and therefore demonstrates that the slip rate of the Alpine Fault is constant along strike. The locking depth for the fault in this region is c. 17 km. Assuming a near vertical dip angle results in unrealistic high slip rates.

Introduction southeast c. 82°) (Barth et al. 2013, Table 1). The Over the last 20 years, the School of Surveying at the vertical component also changes at this point such University of Otago has established a dense geodetic that the Pacific Plate is uplifting north of the Martyr network extending across the South Island of New Zeal- River and the Australian Plate is uplifting to the south. and. The network extends from the east coast near This region has been the subject of geodetic investi- Dunedin across Central Otago and the Alpine Fault to gations since the 1980s (Blick 1986; Pearson 1990) that the west coast near Haast (Figure 1). This profile, were based on the analysis of triangulation measure- which crosses the Central Otago and South Westland ments, but subject to large measurement errors. The regions, is located in the southern half of the South advent of satellite geodesy resulted in improved Island of New Zealand. The eastern half of the profile measurement precision that enabled studies such as is dominated by the Otago fault system, which is charac- Pearson et al. (2000) to examine slip rates on the terised by actively growing asymmetric anticlines above Alpine Fault with a best fit model that accommodates buried reverse faults (Beanland and Berryman 1989; c. 75% of the relative plate motion and a locking Jackson et al. 1996; Litchfield and Norris 2000; Litchfield depth of c. 10 km. For the central South Island, Beavan 2001). These structures are periodically active followed et al. (1999) showed that the majority of the observed by long periods of quiescence when the activity migrates velocity signal (50–70%) is uniform slip along strike to another structure in the region (Beanland and Berry- of the Southern Alps with a shallower locking depth man 1989; Litchfield and Norris 2000). of 5–8 km, which is consistent with higher crustal Farther west, the tectonics are dominated by the temperatures associated with a thinner crust. On the plate boundary zone where, in the central South Island, eastern side of the Southern Alps and away from the boundary takes up oblique convergence that tran- the Alpine Fault, Denys et al. (2014, 2016) showed sitions in the southwestern South Island (South West- the spatial variation in strain accumulation within the land to Fiordland) to subduction of the Australian Otago fault system. Plate (Wallace et al. 2007). Barth et al. (2013) identify In addition to the Otago geodetic data (Denys et al. Martyr River as the change point between the central 2014, 2016), this study includes recent geodetic and southern Alpine Fault. North of Martyr River the measurements from sites across the Haast Pass that plate motion is accommodated as oblique strike-slip have not been measured for many years and data (strike 55°, dipping southeast c. 45°), whereas south from a recently established network, the Cascade the motion is almost pure strike-slip and the Alpine array that extends from the Arawhata River to Lake Fault becomes nearly vertical (strike 52°, dipping McKerrow, a region that previously had very few

CONTACT Paul H. Denys [email protected] Supplemental data for this article can be accessed here https://doi.org/10.1080/00288306.2018.1494006. © 2018 The Royal Society of New Zealand 2 C. J. PAGE ET AL.

phase centre models were significantly less accurate than for more modern antenna. To overcome the centring and antenna height error inherent in traditional GPS campaigns, we have established two networks composed of force-centred marks. In these networks, each antenna is connected to a fixed height adaptor that is attached directly to a 5/8′′ threaded rod epoxied into rock. The height of the antenna reference point above the ground depends on the length of the fixed height adaptor and would generally range between 0.055 and 0.15 m. Secure sites are chosen so that equipment can be left unattended for long periods. The two networks containing force-centred points are the Cascade array consisting of marks epoxied into rock outcrops located on ridge tops above the bush line. The array, which was established in 2012, is accessible only by helicopter and observed at c. 1-year intervals between 2013 and 2016. The Cen- tral Otago network is similar to the Cascade array Figure 1. Network sites. ( ) Continuous GNSS sites, ( ) cam- except that the marks were developed so that each paign sites and ( ) new Cascade array campaign sites. Big site had easy, all-weather access (two-wheeled drive Bay profile sites have black borders (see profiles G and H) vehicle) as well as being secure so that the equipment using sites south of Martyr River (Table 2, Models G and H). could be left unattended. The first sequence of Central Otago marks was established in 2004, a second phase in geodetic measurements. Together, these data form a late 2005 and a third phase in 2009. The mark distri- broad profile between Dunedin to Haast, and south bution is governed by the layout of the road network, of Jackson Bay (Figure 1). which in turn tends to follow the valleys and basins of the region. This allows for easy and fast access to the marks, but results in the mark distribution being Campaign GPS biased towards lower elevations. The surveys have Our study incorporates campaign and continuous been conducted using a reasonably consistent set of measurements collected over the past 20 years. These equipment. All measurements since 2004 have used data have been observed using traditional GPS Trimble 5700, R7 and R10 receivers and Trimble Geo- methods: tripod, tribrach and antenna set up on tra- detic, Trimble Geodetic 2 or R10 antennas. ditional surveying marks, typically stainless pins with centring holes grouted into rock or set in concrete. GPS processing Nominally observing sessions are 2 days (48 h) long, although some of the earlier campaigns were observed GPS data have been processed using the Bernese soft- with sessions of < 24 h. Although this method is versa- ware package (v. 5.2) (Dach et al. 2015) using 24 h tile and expedient, it is prone to centring and antenna daily position solutions. The Centre for Orbit Determi- height measurement errors. The use of different equip- nation (CODE) precise satellite orbit and clock par- ment (e.g. antenna) between campaigns, can result in ameters, together with the I08.ATX absolute GPS positional errors. This is particularly true for the receiver and satellite antenna phase centre model early measurements in the 1990s when the antenna (Schmid et al. 2007) are used to generate daily position

Table 1. Model parameters for the single- and double-fault models estimated using both the pre- and post-Dusky Sound 2009 (DS2009) velocity ﬁelds. Note that a positive slip rate implies dextral motion. The dip of the Alpine Fault is assumed to be 55° and the antithetic fault is assumed to be 46°.

Fault model Fault(s) Strike slip rate Locking depth mm/yr ± 1σ mm/yr km ± 1σ km A Pre-DS2009 Alpine Fault 1 35.0 0.4 21 1 B Pre-DS2009 Alpine Fault 2 30.0 1.5 17 1 Antithetic fault 4.2 1.0 17 Assumed C Post-DS2009 Alpine Fault 1 37.1 0.8 11 2 H Post-DS2009 1 36.8 0.2 20 fixed Assumed Alpine Fault (Big Bay profile) NEW ZEALAND JOURNAL OF GEOLOGY AND GEOPHYSICS 3 time series. The carrier phase ionosphere-free linear decay. Following Denys and Pearson (2015, 2016), we combination is used to correct the first-order iono- model this displacement by modifying Equation (1) sphere. Higher-order ionospheric effects are not con- such that: sidered here because Hernández-Pajares et al. (2007) X(t) = X + vX(t − t ) and Petrie et al. (2010) showed that these effects are 0 0 ff X(t) = X0 + vX(t − t0) < 1 mm. Tropospheric e ects are modelled using the (3) Global Mapping Function, which maps the zenith tro- n t − t t , tk + k OC + A k k= k k log posphere delay to the elevation of each observation 1 tk t ≥ tk (Boehm et al. 2007). A 10° elevation cut-off angle is OC ff A used, a compromise to constrain tropospheric effects where k is the coseismic o set (m), k is the amplitude but minimise multipath errors. Non-tidal atmospheric of the post-seismic decay (m), tk is the decay time scale loading displacements are modelled according to Ray (yr), tk the event time (yr) and nk is the number of and Ponte (2003). The effects of ocean loading are cor- events. Modelling of seasonal effects (annual and semi- rected using the FES2004 model (Lyard et al. 2006) from annual cyclic terms) and including any position offsets the Onsala Space Observatory (holt.oso.chalmers.se/ caused by equipment changes (e.g. replacement anten- loading). Ambiguity resolution involves a recursive nas) is also included in the continuous position time strategy that includes code and phase-based wide lane, series (see Denys and Pearson 2015, 2016 for details). QIF and direct L1/L2 fixed ambiguities, depending Figure 2 shows the velocity vectors in a Pacific Plate upon baseline length. The ITRF2014 reference frame is fixed frame after removing the MORVEL (Pacific) plate realised through the Helmert three-parameter trans- motion (DeMets et al. 2010). The velocity vectors are formation of the daily coordinate positions. Global generally consistent at the few millimetre level on the IGS sites include those on the stable Pacific, Antarctic east coast, but increase rapidly towards the Alpine and Australian plates. Data outliers were removed Fault. As noted by Sutherland et al. (2006), the vectors using the Median Absolute Deviation (MAD) robust in this region are close to being parallel to the trace of estimator at the 4σ level with σ =1.4826×MAD. the Alpine Fault, which would imply that the fault is predominantly strike-slip.

Velocity estimation Modelling infinite faults We model the site position time series assuming linear velocities. Hence the basic equation is: We model the fault parallel and normal velocity profiles using equations for infinite faults developed X(t) = X0 + vX(t − t0) (1) where X(t) is the position (ordinate) (m) at time t (yr), X0 is the reference position (m) at reference time t0 and v M X is the site velocity (m/yr). The W 7.8 Dusky Sound 2009 earthquake (Beavan et al. 2010) not only caused earth displacement at the time of the event throughout the lower South Island (Crook and Donnelly 2013), but also resulted in a new velocity field since the time of the event. Equation (1) is modified to include transient velocity events that act as a time dependent offset (i.e. ramp) with:

X(t) = X0 + vX(t − t0) n + l OC + v t − t [ l Vl ( l )] (2) l=1 1, 0

t1,l , t , t2,l OC ff v where l is a coseismic o set (m), Vl is the transient t t velocity (m/yr), 1,l is the start time and 2,l is the end t time of the velocity event ( 2,l may be the end of the time series if the velocity change is ongoing as is the case for the Dusky Sound post-seismic relaxation) and nlis the number of events. Figure 2. Site velocities relative to the MORVEL Pacific Plate In the case of the continuous GNSS sites where the pole (DeMets et al. 2010). ( ) Pre-Dusky Sound 2009 earth- position time series are estimated daily, the post-seis- quake velocities and ( ) post-Dusky Sound 2009 velocities. mic deformation is modelled as a characteristic log Error ellipses are shown as 95% confidence intervals. 4 C. J. PAGE ET AL. by Savage (1983) and similar to the procedure reason, we analysed the pre- and post-earthquake described by Beavan et al. (1999) and Pearson et al. velocities separately. We first inverted the pre-earth- (2000). We assume uniformity along the fault, thus quake data by considering the straightforward reducing the problem to two dimensions. This simple single-fault model in which all deformation is but effective model assumes a locking depth below assumed to be slip on the Alpine Fault. The fault- which the fault slips freely at a uniform slip rate. In parallel component of motion is shown in Figure 3 this case, the plate boundary parallel component of vel- and the model parameters are summarised in Table ocity, VP, is: 1 (Model A). The single-fault model with a dip of 55° is a reasonable fit except that the predicted S x − d tand V = V + −1 fault-parallel profile is systematically too low in the p 0 p tan d range 30–60 km, resulting in a standard error of ff where V is the slip-rate offset (mm/yr), S is the slip unit weight (SEUW) of 1.9. This systematic di erence 0 fl rate (mm/yr), x is the perpendicular distance from is likely to indicate the in uence of distributed defor- the fault (km), d is the locking depth of the fault mation and/or other unmodelled faults. (km) and d is the dip of the fault (°). (There is a similar Previous studies (Beavan et al. 1999; Pearson et al. equation for the plate boundary normal component of 2000) suggest that this extra deformation can be mod- the velocity field, VN , however we have included only elled as a second structure running parallel to the the plate boundary parallel component in our model- Alpine Fault but dipping to the west. Using a block fi ling because the velocity vectors shown in Figure 2 model, Wallace et al. (2007) also showed that the t indicate that this portion of the Alpine Fault is predo- of the GPS data improved with a discrete boundary minately strike-slip.) c. 150 km east of the Alpine Fault that accommodated – As a first step towards a tectonic interpretation of c. 3 4 mm/yr of deformation. Alternatively, the extra the velocity field, we follow the procedure outlined by deformation could be ascribed to distributed defor- Pearson et al. (2000) and constructed profiles of the mation within the Southern Alps. Following Pearson fi fault parallel and fault normal components of the vel- et al. (2000), we modelled the pro le with two oppo- fi ocity as a function of distance from the Alpine Fault. sitely dipping shear zones, one of which is identi ed We inverted the velocity data to obtain fault slip rates as the Alpine Fault , whereas the second, west-dipping by linearising Equation (4) which models slip on fault is located c. 150 km east of the Alpine Fault as a infinite faults (Savage 1983) and use standard least proxy for distributed deformation associated with square techniques to find a solution that minimised other structures. Equation (5) illustrates the double- the sum-squared of the weighted residuals. fault model in which the right-hand term is the second d D Applying this technique requires that the geometry fault with dip, 2, and represents perpendicular dis- of the Alpine Fault is known. Our study area is located tance between the double-faults. in the transition zone between double-fault geometries; s x − d tand 1 −1 1 namely, the northern section extending from Haast to VP = V + 0 p tan d Arthur’s Pass (central South Island), where the fault has a dip of c. 50° (Sibson et al. 1979; Norris and s − ()−D − x d tand2 + 2 tan 1 Cooper 2001), and the southern section (South West- p d land) where the Alpine Fault has a nearly vertical dip fi (Sutherland and Norris 1995; Barth et al. 2013). The parameters of our best- tting double-fault Although the surface dip of the Alpine Fault is well model are listed in Table 1 (Model B), whereas established by geological studies, the dip of the plate Figure 4 shows the fault-parallel velocities. Incorpor- interface at depth is controversial. For example, Lamb ating the second fault plane 150 km east of the fi and Smith (2013) model the Alpine Fault/Puysegur Alpine Fault produces an improvement in t with subduction zone as having a fairly consistent dip of the SEUW decreasing to 1.6. The locking depth c. 50°. We explored the effect of differing fault geome- and slip rate of the Alpine Fault from our single- fi tries by developing profiles with fault dip of between fault model are signi cantly greater than those from 45° and 80°. Our preferred fault dip is 55° based on Pearson et al. (2000). fi the geological evidence from Barth et al. (2013) show- Figure 5 shows the fault-parallel velocity pro le for a ing that the dip of the Alpine Fault steepens in our single-fault model based on only the post-earthquake ff study area. velocities (Table 1, Model C). A major di erence is the significant gradient of the fault-parallel velocities as a result of the Cascade array measurements. The Results steepness is not obvious with the more limited data Denys et al. (2016) showed that there is a significant set for the pre-Dusky Sound 2009 velocity field. A change in the secular velocity field in Central Otago second difference between the pre- and post-earth- due to the 2009 Dusky Sound earthquake. For this quake profiles is the lack of measurements between 6 NEW ZEALAND JOURNAL OF GEOLOGY AND GEOPHYSICS 5

Figure 3. Single-fault parallel component of the velocity field based on the pre-Dusky Sound 2009 velocities (Model A). The solid line shows predicted velocity for an infinite Alpine Fault. Triangles represent observed velocities and error bars are at the 1σ level. The x-axis shows distance from the surface trace of the Alpine Fault in km. and 88 km from the Alpine Fault. Because these data single-fault models for the pre- and post-Dusky are not available, it is not possible to invert for a two- Sound 2009 earthquake data sets, but the locking fault model because it is the data in this region that depth is clearly much less for the post-Dusky Sound constrain the second fault. earthquake velocity field. This difference shows the A comparison of the fault parameters in Table 1 importance of the measurements between 6 and shows that the slip rates are very similar between the 88 km in controlling the locking depth.

Figure 4. Double-fault parallel component of the velocity ﬁeld based on the pre-Dusky Sound 2009 velocities (Model B). The solid line shows predicted velocity for an inﬁnite Alpine Fault and a second parallel antithetic fault located 152 km southeast (i.e. inboard). Triangles represent observed velocities and error bars are at the 1σ level. The x-axis shows distance from the surface trace of the Alpine Fault in km.

Figure 5. Single-fault parallel component of the velocity ﬁeld based on the post-Dusky Sound 2009 velocities (Model C). The solid line shows predicted velocity for an inﬁnite Alpine Fault. Triangles represent observed velocities and error bars are at the 1σ errors. 6 C. J. PAGE ET AL.

Table 2. Model parameters for single and double-fault models estimated using both the pre- and post-Dusky Sound 2009 (DS2009) velocity fields assuming a near vertical Alpine Fault (80° dip) and the antithetic fault is assumed to be 46°. Fault model fault(s) Strike slip rate Locking depth mm/yr ± 1σ mm/yr km ± 1σ km D Pre-DS2009 Alpine Fault 1 46.4 0.2 20 fixed Assumed E Pre-DS2009 Alpine Fault 2 47.6 0.6 20 fixed Assumed Antithetic fault 2.0 0.4 20 fixed Assumed F Post-DS2009 Alpine Fault 1 47.6 0.5 20 fixed Assumed G Post-DS2009 Alpine Fault (Big Bay profile) 1 45.7 0.5 20 fixed Assumed

The single-fault models (Models A, C) give slip measurement errors will decrease allowing for a more rates on the Alpine Fault that are slightly higher definitive test to be made. and locking depths that encompass those of Wallace Our study shows that profiles (Table 2, Models D–F) et al. (2007) of 30 mm/yr and 18 km, respectively. that assume a vertical (80°) dip for the Alpine Fault However, the double-fault model (Model B) results provide a good fit to the measured GNSS velocities. in a slip rate and a locking depth that that are com- Indeed, the profiles are virtually indistinguishable parable. The similarity between the two studies is not from those shown in Figures 3–6. However, the slip surprising because the Wallace et al. (2007) model is rates from the inversions are significantly higher. dominated by the Alpine Fault in this part of the Indeed, the resulting slip rates are equal to the total South Island and therefore is one-dimensional in relative plate motion between the Pacific and Australia nature. However, the two studies differ in that our plates for this part of the plate boundary (40 mm/yr models are constrained by additional GNSS data based on the MORVEL plate model, DeMets et al. that were not available to Wallace et al. (2007), par- 2010). These values seem to be unreasonably high, particularly in South Westland. Our slip rates are higher ticularly given that other faults are known to be active than those of Beavan et al. (1999) and Pearson et al. within Otago, on the inboard or the Pacific Plate side of (2000) for the double-fault model, but are quite simi- the Alpine Fault (Beanland and Berryman 1989; Jack- lar for the strike-slip rates from their single-fault son et al. 1996; Litchfield and Norris 2000). For this model. The strike-slip rate from our double-fault reason, we prefer the model with the fault dipping at model (Model B) accommodates c. 75% of the MOR- 55°. VEL relative PA/AU plate motion. Barth et al. (2013) suggested that the dip of the One of the questions that the Cascade array was Alpine Fault changes along strike in South Westland designed to answer was whether there was any evi- with the section south of the Martyr River having a dence for partial locking on the Alpine Fault. Figure 6 near vertical dip. However, it is not clear whether the shows the velocity data within ± 10 km of the surface dip in Barth et al. (2013), which relates to fault geome- trace of the Alpine Fault. If partial locking was appar- try in the near surface, is representative of the fault geo- ent, we would expect to see a discontinuity in the metry at mid crustal depths. To test this hypothesis, we measurements. Because all of the points appear to fit developed a revised profile centred on Big Bay (44.25°S, the predicted curve (solid line) within the estimated 168.05°E, sites with a black border, Figure 1) that incor- uncertainties, there does not seem to be any evidence porates only points south of the Martyr River (Table 2, for partial locking at the present time. With more Model G). Inverting the data gives a slip rate of 46 mm/ time and measurements, it is expected that the yr for the Alpine Fault, whereas a profile assuming a

Figure 6. Fault parallel component of the velocity field for the region within 10 km of the surface trace of the Alpine Fault. The solid line shows predicted velocity for an infinite Alpine Fault. Triangles represent observed velocities and error bars are at the 1σ errors. NEW ZEALAND JOURNAL OF GEOLOGY AND GEOPHYSICS 7 dip of 55° (consistent with the rest of the Alpine Fault that are greater than the total relative plate motion see Table 1, Model H) gives a slip rate of 37 mm/yr. between the Pacific and Australia plates. For this Note that both these inversions used velocities derived reason, we prefer a model in which the dip of the only from post-Dusky Sound 2009 data and in both Alpine Fault at mid crustal depths is closer to the 50° cases we constrained the locking depth to 20 km that characterises the Alpine Fault farther north. because there were insufficient velocity measurements between 10 and 80 km to allow us to invert for this parameter. Also, the model with 55° dip has an SEUW of Acknowledgements 1.0 compared with 1.1 for the vertical fault, which indi- fi Fieldwork support undertaken by Mike Denham and Alas- cates that the 55° dipping model has a slightly better t tair Neaves (School of Surveying) together with numerous to the data. Otago University students. We thank GeoNet (www. It should be noted that the modelling procedure geonet.org.nz), the New Zealand Earthquake Commission implicitly assumes an infinite fault without any along and Land Information New Zealand (LINZ) for the oper- fi strike-slip change in geometry. For this reason, a ation and nancial support of the cGNSS network. We definitive study of along strike variation in the geome- thank Laura Wallace, an anonymous reviewer and editor ffi Phaedra Upton for their reviews and useful comments that try and locking coe cient will require more sophisti- have improved the quality of this paper. cated modelling procedures that have the capability for three-dimensional variation of fault geometry. Disclosure statement Conclusions No potential conflict of interest was reported by the authors. The Alpine Fault is a major strike-slip fault that accommodates c. 75% of the current plate motion. Funding It is well recognised that large earthquakes have occurred at regular intervals over geological time The initial establishment of this Central Otago deformation making this system important to understand (see network was partly funded by an Earthquake Commission grant. The Cascade array was funded through two Otago e.g. Sutherland et al. 2007). Although the Alpine University research grants. All other GPS data has been Fault has been researched extensively, this paper funded by NZ Ministry for Science and Innovation (formerly includes new GPS data from South Westland. the Foundation for Research, Science and Technology) Because of the effect of the post-seismic deformation through the GNS Science New Zealand Geological Hazards from the 2009 Dusky Sound earthquake, the velocity and Society Programme (C05X0804) with contributions to field has been divided into pre- and post-earthquake the Impacts of Global Plate Tectonics in and around New Zealand Programme (C05X02023). event data sets. For this broad profile, the Alpine Fault is initially modelled as a single (infinitely long) fault that is uniformly slipping below a (con- References stant) locking depth. Both models have similar strike-slip rates of 37–35 mm/yr respectively. By con- Barth NC, Boulton C, Carpenter BM, Batt GE, Toy VG. 2013. Slip localization on the southern Alpine fault, New trast, the locking depths are significantly different for Zealand. Tectonics. 32(3):620–640. the two models at 21 km (pre-Dusky Sound 2009 Beanland S, Berryman KR. 1989. Style and episodicity of late earthquake) and 11 km (post earthquake). However, Quaternary activity on the Pisa-Grandview fault zone, the models systematically predict lower strike-slip central Otago, New Zealand. New Zealand Journal of rates that are interpreted to represent distributed Geology and Geophysics. 32:451–461. deformation to the east of the Alpine Fault. Beavan J, Moore M, Pearson C, Henderson M, Parsons B, Bourne S, England P, Walcott D, Blick G, Darby D, To account for the (inboard) distributed defor- Hodgkinson K. 1999. Crustal deformation during 1994– mation, a second (proxy) fault is modelled some 1998 due to oblique continental collision in the central 150 km east of the Alpine Fault as a double-fault southern Alps, New Zealand, and implications for seismic model. The strike-slip rate reduces to 30 mm/yr, with potential of the Alpine fault. Journal of Geophysical – a locking depth of 17 km in this model compared Research: Solid Earth. 104(B11):25233 25255. Beavan J, Samsonov K, Denys P, Sutherland R, Palmer N, with the single-fault case. Our slip rates are signifi- Denham M. 2010. Oblique slip on the Puysegur subduc- cantly greater than those given by Sutherland et al. tion interface in the 2009 July MW 7.8 Dusky Sound earth- (2006). At this stage, the double-fault model can be quake from GPS and InSAR observations: implications for estimated only using the pre-Dusky Sound velocity the tectonics of southwestern New Zealand. Geophysical field. More GPS measurements are required in the Journal International. 183:1265–1286. profile between c. 10 and 80 km east of the Alpine Blick GH. 1986. Geodetic determination of crustal strain from old survey data in Central Otago. In: Reilly WI, Fault. Harford BE, editor. Recent crustal movements of the Our data can also be modelled using a vertical pacific region. Wellington: Royal Society of New Alpine Fault, however, this model produces slip rates Zealand, Bulletin 24; p. 47–54. 8 C. J. PAGE ET AL.

Boehm J, Heinkelmann R, Schuh H. 2007. Short note: a glo- Lyard F, Lefevre F, Letellier T, Francis O. 2006. Modelling bal model of pressure and temperature for geodetic appli- the global ocean tides: modern insights from FES2004. cations. Journal of Geodesy. 81(10):679–683. Ocean Dynamics. 56(5-6):394–415. Crook C, Donnelly N. 2013. Updating the NZGD2000 defor- Norris RJ, Cooper AF. 2001. Late Quaternary slip rates and mation model, Proceedings of the 125th Annual slip partitioning on the Alpine fault, New Zealand. Conference of the New Zealand Institute of Surveyors Journal of Structural Geology. 23(2):507–520. (NZIS), August 28–31, 2013; Dunedin, New Zealand. Pearson C. 1990. Extent and tectonic significance of the Dach R, Lutz S, Walser P, Fridez P. 2015. Bernese GNSS soft- Central Otago shear strain anomaly. New Zealand ware version 5.2. Bern: Astronomical Institute, University Journal of Geology and Geophysics. 33:295–301. of Bern. Pearson C, Denys P, Hodgkinson K. 2000. Geodetic con- DeMets C, Gordon RG, Argus DF. 2010. Geologically cur- straints on the kinematics of the Alpine fault in the rent plate motions. Geophysical Journal International. southern South Island of New Zealand, using results 181(1):1–80. doi:10.1111/j.1365-246X.2009.04491.x; see from the Hawea-Haast GPS Transect. Geophysical also Erratum, Geophysical Journal International, Research Letters. 27(9):1319–1322. doi:10.1111/j.1365-246X.2011.05186.x, 2011. Petrie EJ, King MA, Moore P, Lavallée DA. 2010. Higher- Denys P, Norris R, Pearson C, Denham M. 2014. A geodetic order ionospheric effects on the GPS reference frame study of the Otago fault system of the South Island of and velocities. Journal of Geophysical Research. 115: New Zealand. In: Rizos C, Willis P., editor. Earth on the B03417. doi:10.1029/2009JB006677. edge: science for a sustainable planet, international Ray RD, Ponte RM. 2003. Barometric tides from ECMWF association of geodesy symposia 139. Springer-Verlag operational analyses. Annales Geophysicae. 21(8):1897– Berlin Heidelberg; p. 151–158. doi:10.1007/978-3-642- 1910. 37222-3_19. Savage J. 1983. Strain accumulation in western United Denys P, Pearson C. 2015. Modelling time dependent transi- States. Annual Review of Earth and Planetary Sciences. ent deformation in New Zealand. Proceedings of 11(1):11–41. International Symposium on GNSS (IS-GNSS 2015), Schmid R, Steigenberger P, Gendt G, Ge M, Rothacher M. November 16–19, 2015; Kyoto, Japan. 2007. Generation of a consistent absolute phase-center Denys P, Pearson C. 2016. Positioning in active deformation correction model for GPS receiver and satellite antennas. zones – implications for NetworkRTK and GNSS proces- Journal of Geodesy. 81(12):781–798. sing engines. FIG Working Week 2016, Recovery from Sibson RH, White SH, Atkinson BK. 1979. Fault rock distri- Disaster. 2–6 May 2016, Christchurch, New Zealand. bution and structure within the Alpine fault zone: a pre- Denys P, Pearson C, Norris R, Denham M. 2016. A geodetic liminary account. Bulletin of the Royal Society of New study of Otago: results of the central Otago deformation Zealand. 18:55–65. network 2004–2014. New Zealand Journal of Geology Sutherland R, Berryman K, Norris R. 2006. Quaternary slip and Geophysics. 59(1):147–156. rate and geomorphology of the Alpine fault: implications Hernández Pajares M, Juan JM, Sanz J, Orús R. 2007. Second for kinematics and seismic hazard in southwest New order ionospheric term in GPS: implementation and Zealand. Geological Society of America Bulletin. 118(3- impact on geodetic estimates. Journal of Geophysical 4):464–474. Research. 112(B8):921. Sutherland R, Eberhart-Phillips D, Harris RA, Stern T, Jackson J, Norris R, Youngson J. 1996. The structural evol- Beavan J, Ellis S, Henrys S, Cox S, Norris RJ, Berryman ution of active fault and fold systems in central Otago, KR, et al. 2007. Do great earthquakes occur on the New Zealand: evidence revealed by drainage patterns. Alpine fault in Central South Island, New Zealand? In: Journal of Structural Geology. 18:217–234. Okaya D, Stern T, Davey F, editors. A continental plate Lamb S, Smith E. 2013. The nature of the plate interface and boundary: tectonics at South Island, New Zealand. driving force of interseismic deformation in the New doi:10.1029/175GM12. Zealand plate-boundary zone, revealed by the continuous Sutherland R, Norris RJ. 1995. Late Quaternary displacement GPS velocity field. Journal of Geophysical Research: Solid rate, paleoseismicity, and geomorphic evolution of the Earth. 118(6):3160–3189. Alpine fault: evidence from Hokuri Creek, South Litchfield NJ. 2001. The titri fault system: Quaternary-active Westland, New Zealand. New Zealand Journal of faults near the leading edge of the Otago reverse fault pro- Geology and Geophysics. 38(4):419–430. vince. New Zealand Journal of Geology and Geophysics. Wallace LM, Beavan J, McCaffrey R, Berryman K, Denys P. 44:517–534. 2007. Balancing the plate motion budget in the South Litchfield NJ, Norris RJ. 2000. Holocene motion on the Island, New Zealand using GPS, geological and seismolo- Akatore fault, South Otago coast, New Zealand. New gical data. Geophysical Journal International. 168 Zealand Journal of Geology and Geophysics. 43:405–418. (1):332–352.